Added option to disable pack buffer memory pools.

Details:
- Added a new configure option, --[en|dis]able-packbuf-pools, which will
  enable or disable the use of internal memory pools for managing buffers
  used for packing. When disabled, the function specified by the cpp
  macro BLIS_MALLOC_POOL is called whenever a packing buffer is needed
  (and BLIS_FREE_POOL is called when the buffer is ready to be released,
  usually at the end of a loop). When enabled, which was the status quo
  prior to this commit, a memory pool data structure is created and
  managed to provide threads with packing buffers. The memory pool
  minimizes calls to bli_malloc_pool() (i.e., the wrapper that calls
  BLIS_MALLOC_POOL), but does so through a somewhat more complex
  mechanism that may incur additional overhead in some (but not all)
  situations. The new option defaults to --enable-packbuf-pools.
- Removed the reinitialization of the memory pools from the level-3
  front-ends and replaced it with automatic reinitialization within the
  pool API's implementation. This required an extra argument to
  bli_pool_checkout_block() in the form of a requested size, but hides
  the complexity entirely from BLIS. And since bli_pool_checkout_block()
  is only ever called within a critical section, this change fixes a
  potential race condition in which threads using contexts with different
  cache blocksizes--most likely a heterogeneous environment--can check
  out pool blocks that are too small for the submatrices it wishes to
  pack. Thanks to Nisanth Padinharepatt for reporting this potential
  issue.
- Removed several functions in light of the relocation of pool reinit,
  including bli_membrk_reinit_pools(), bli_memsys_reinit(),
  bli_pool_reinit_if(), and bli_check_requested_block_size_for_pool().
- Updated the testsuite to print whether the memory pools are enabled or
  disabled.

Author: fgvanzee

build/bli_config.h.in

@@ -52,6 +52,10 @@
 #define BLIS_ENABLE_PTHREADS
 #endif
 
+#if @enable_packbuf_pools@
+#define BLIS_ENABLE_PACKBUF_POOLS
+#endif
+
 #if @int_type_size@ == 64
 #define BLIS_INT_TYPE_SIZE 64
 #elif @int_type_size@ == 32

configure

@@ -99,6 +99,21 @@ print_usage()
 	echo "                 --disable-threading is specified, threading will be"
 	echo "                 disabled. The default is 'no'."
 	echo " "
+	echo "   --disable-packbuf-pools, --enable-packbuf-pools"
+	echo " "
+	echo "                 Disable (enabled by default) use of internal memory"
+	echo "                 pools for managing packing buffers. When disabled,"
+	echo "                 the function specified by BLIS_MALLOC_POOL is called"
+	echo "                 on-demand, whenever a packing buffer is needed, and"
+	echo "                 the buffer is released via the function specified by"
+	echo "                 BLIS_FREE_POOL() when the loop in which it was"
+	echo "                 allocated terminates. When enabled, the memory pools"
+	echo "                 minimize calls to both BLIS_MALLOC_POOL() and"
+	echo "                 BLIS_FREE_POOL(), especially in a multithreaded"
+	echo "                 environment, but does so through a mechanism that may"
+	echo "                 incur additional overhead in some (but not all)"
+	echo "                 situations."
+	echo " "
 	echo "   -q, --quiet   Suppress informational output. By default, configure"
 	echo "                 is verbose. (NOTE: -q is not yet implemented)"
 	echo " "
@@ -502,6 +517,7 @@ main()
 	enable_verbose='no'
 	enable_static='yes'
 	enable_shared='no'
+	enable_packbuf_pools='yes'
 	int_type_size=0
 	blas2blis_int_type_size=32
 	enable_blas2blis='yes'
@@ -582,6 +598,12 @@ main()
 					disable-threading)
 						threading_model='no'
 						;;
+					enable-packbuf-pools)
+						enable_packbuf_pools='yes'
+						;;
+					disable-packbuf-pools)
+						enable_packbuf_pools='no'
+						;;
 					int-size=*)
 						int_type_size=${OPTARG#*=}
 						;;
@@ -1029,6 +1051,13 @@ main()
 	fi
 
 	# Convert 'yes' and 'no' flags to booleans.
+	if [ "x${enable_packbuf_pools}" = "xyes" ]; then
+		echo "${script_name}: internal memory pools for packing buffers are enabled."
+		enable_packbuf_pools_01=1
+	else
+		echo "${script_name}: internal memory pools for packing buffers are disabled."
+		enable_packbuf_pools_01=0
+	fi
 	if [ "x${enable_blas2blis}" = "xyes" ]; then
 		echo "${script_name}: the BLAS compatibility layer is enabled."
 		enable_blas2blis_01=1
@@ -1135,6 +1164,7 @@ main()
 		| perl -pe "s/\@kernel_list_defines\@/${kernel_list_defines}/g" \
 		| sed   -e "s/@enable_openmp@/${enable_openmp_01}/g" \
 		| sed   -e "s/@enable_pthreads@/${enable_pthreads_01}/g" \
+		| sed   -e "s/@enable_packbuf_pools@/${enable_packbuf_pools_01}/g" \
 		| sed   -e "s/@int_type_size@/${int_type_size}/g" \
 		| sed   -e "s/@blas2blis_int_type_size@/${blas2blis_int_type_size}/g" \
 		| sed   -e "s/@enable_blas2blis@/${enable_blas2blis_01}/g" \

frame/3/gemm/bli_gemm_front.c

@@ -62,10 +62,6 @@ void bli_gemm_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A, B, and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );

frame/3/hemm/bli_hemm_front.c

@@ -63,10 +63,6 @@ void bli_hemm_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A, B, and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );

frame/3/her2k/bli_her2k_front.c

@@ -67,10 +67,6 @@ void bli_her2k_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A, B, and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );

frame/3/herk/bli_herk_front.c

@@ -63,10 +63,6 @@ void bli_herk_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *c, c_local );

frame/3/symm/bli_symm_front.c

@@ -63,10 +63,6 @@ void bli_symm_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A, B, and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );

frame/3/syr2k/bli_syr2k_front.c

@@ -64,10 +64,6 @@ void bli_syr2k_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A, B, and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );

frame/3/syrk/bli_syrk_front.c

@@ -61,10 +61,6 @@ void bli_syrk_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A and C in case we need to apply transformations.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *c, c_local );

frame/3/trmm/bli_trmm_front.c

@@ -61,10 +61,6 @@ void bli_trmm_front
 		return;
 	}
 
-	// Reinitialize the memory allocator to accommodate the blocksizes
-	// in the current context.
-	bli_memsys_reinit( cntx );
-
 	// Alias A and B so we can tweak the objects if necessary.
 	bli_obj_alias_to( *a, a_local );
 	bli_obj_alias_to( *b, b_local );